quality-strategy

Quality Strategy Skills

Claude Code skills that interview you to produce a software quality strategy — who matters for your project, what they value, where you're exposed, and what to do about it — plus the test strategy and tooling strategy that turn it into action. When most of your code is written by agents who don't know what quality means for your project, an explicit strategy is what stops them shipping confidently in the wrong direction.

Grounded in Edmund Pringle's quality framework: quality is value to someone who matters; testing is investigation to find out what's actually true; risk is danger to quality. For solo developers, small teams, engineering leaders — and anyone whose codebase agents now work in. You don't need a repo: running it at the idea stage is a first-class use.

Status: alpha. Shared with a first wave of testers. The skills work and have been exercised hard against simulated users, but real-world mileage is limited. Expect rough edges, tell us where it misfires (Feedback), and read Known limitations first — we're not hiding the gaps.

Install

This is a Claude Code plugin:

/plugin marketplace add tollens-ai/quality-strategy-skills
/plugin install quality-strategy@tollens

Then, in any project, start with /quality-strategy. Output goes to quality/strategy.md at the project root. (Skills are also available namespaced, e.g. /quality-strategy:test-strategy, if a bare name ever collides with another plugin.)

New versions ship regularly — run claude plugin update quality-strategy to pick them up; CHANGELOG.md says what changed.

The typical flow

1. /quality-strategy — the main event. A structured interview produces quality/strategy.md, ends with a built-in audit, and points you at the follow-ons.
2. /test-strategy — turns the strategy into an investigation plan: what to find out, in what order, split between humans and agents.
3. /tooling-strategy — when the docs surfaced things you can't measure or judge yet (common, and a finding rather than a failure): a prioritised build plan for the missing oracles and instruments. Steps 2 and 3 swap when the risk map comes out mostly blind — the skills recommend the order themselves; you can always overrule.
4. /strategy-variants (optional) — a one-pager or client-safe version to circulate.
5. /quality-artefacts (optional) — describe the view you want ("a tweetable summary of where quality stands", "a dashboard of the payment risks for my standup") and it designs a bespoke, self-contained SVG/HTML artefact from your strategy — honest about Unknowns, built to be screenshotted and shared. Worked examples in examples/fernly/quality/artefacts/.

How the skills fit together

Five skills you start (bold); the rest are checks and audits the strategies invoke for you as they run (dotted arrows — each also works standalone). Per-skill details: docs/SKILLS.md.

flowchart TD
    QS["<b>/quality-strategy</b><br/>the 7-step interview — START HERE"]
    OA["/oracle-adequacy<br/><i>can the 'where we are' claims<br/>actually be judged?</i>"]
    CC["/contradiction-check<br/><i>internal consistency,<br/>every step boundary</i>"]
    OD["/operational-distillation<br/><i>TL;DR + triage rubric<br/>at the top of the doc</i>"]
    QSR["/quality-strategy-review<br/><i>the closing audit</i>"]
    SDOC[/"quality/strategy.md"/]
    TS["<b>/test-strategy</b><br/>the investigation plan"]
    TA["/tooling-adequacy<br/><i>can each learning need<br/>actually be answered?</i>"]
    TSR["/test-strategy-review<br/><i>the closing audit</i>"]
    TDOC[/"quality/test-strategy.md"/]
    TOOL["<b>/tooling-strategy</b><br/>the oracle/instrument build plan"]
    ODOC[/"quality/tooling-strategy.md"/]
    SV["<b>/strategy-variants</b><br/>audience-facing variants"]
    VDOC[/"one-pager ·<br/>client-safe version"/]
    QA["<b>/quality-artefacts</b><br/>bespoke shareable visuals"]
    ADOC[/"quality/artefacts/<br/>*.svg · *.html"/]
    BUILD(["build the oracles<br/>& instruments"])

    QS -.-> OA
    QS -.-> CC
    QS -.-> OD
    QS -.-> QSR
    QS --> SDOC

    SDOC -->|"risk map mostly<br/>answerable"| TS
    SDOC -->|"risk map mostly blind —<br/>Unknowns + oracle gaps"| TOOL
    SDOC -->|"need something<br/>to circulate"| SV
    SDOC -->|"need something glanceable —<br/>describe the view you want"| QA

    TS -.-> TA
    TS -.-> TSR
    TS --> TDOC
    TDOC -->|"blocked learning needs —<br/>the sharpened demand"| TOOL

    TOOL --> ODOC
    ODOC --> BUILD
    BUILD -.->|"Unknowns become knowable —<br/>update the risk map,<br/>unblock the learning needs"| SDOC

    SV --> VDOC
    QA --> ADOC